Artificial intelligence cannot replace human respondents in opinion polls, at least for now.

Analysis

The claim gains traction because AI promises cheaper, faster data collection, and pollsters are under pressure from declining response rates and rising costs. However, the Pew study shows that current large‑language models miss recent political shifts, rely on outdated training data, and fall back on demographic stereotypes, producing results that would mislead readers if presented as real public opinion. People may overlook these shortcomings when headlines highlight speed or cost savings, assuming AI can perfectly mimic human judgment. The nuance is that while AI can assist in tasks like coding open‑ended responses, it cannot yet replace the lived experience and contextual understanding of real respondents. Misrepresenting AI‑generated answers as genuine public sentiment could distort policy debates, election forecasts, and public discourse, underscoring why the verdict remains that AI is not a viable replacement for human polling at this time.